R ENDICONTI
del
S EMINARIO M ATEMATICO
della
U NIVERSITÀ DI P ADOVA
J. C HABROWSKI
Dirichlet problem for a linear elliptic equation in unbounded domains with L
2-boundary data
Rendiconti del Seminario Matematico della Università di Padova, tome 71 (1984), p. 287-328
<http://www.numdam.org/item?id=RSMUP_1984__71__287_0>
© Rendiconti del Seminario Matematico della Università di Padova, 1984, tous droits réservés.
L’accès aux archives de la revue « Rendiconti del Seminario Matematico della Università di Padova » (http://rendiconti.math.unipd.it/) implique l’accord avec les conditions générales d’utilisation (http://www.numdam.org/conditions).
Toute utilisation commerciale ou impression systématique est constitutive d’une infraction pénale. Toute copie ou impression de ce fichier doit conte- nir la présente mention de copyright.
Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques
http://www.numdam.org/
Elliptic Equation
in Unbounded Domains with L2 - Boundary Data.
J. CHABROWSKI
1. Introduction.
The main purposes of this paper are to
investigate
the Dirichletproblem
for theelliptic equation
in a
half-space
and acomplement
of a bounded open set. We shall refer to the secondproblem
as the exterior Dirichletproblem.
Given an open set Q c we denote
by
the Banach space of functions u inZ2(S~) having
weak(distributional)
derivativesDiu.
(i = 1, ..., n)
inL2(Q).
A norm is introducedby defining
The closure of in
Wl,2(Q)
is denotedby
A local space consists of functionsbelonging
to for every bounded open set ,~’ such that S~.To motivate our
approach
to the Dirichletproblem
assume forsimplicity
that .~ isuniformly elliptic
and the coefficients c(*)
Indirizzo dell’A.:Department
of Mathematics,University
of Queens- land, St. Lucia Queensland 4067, Australia.and
f
are measurable and bounded on SZ. A function u is said to bea weak solution of the
equation (1)
if u E and u satisfiesfor every v
EWl,2(Q)
withcompact support
in ,~. Let q E_L2(aS2)
and assume that there is a function ffJl E
Wl,2(Q)
such that q = ffJ1on a,~ in the sense of trace. A weak solution in of the equa- tion
(1)
is a solution of the Dirichletproblem
with theboundary
condi-tion u = 99 on 8Q if The basic results
concerning
the Dirichlet
problem
in W1,2-framework can be found inLadyzhens- kaja
and Ural’ceva[16], Gilbarg
andTrudinger [10]
andStampacchia [24], [25].
In the above definition it is assumed that theboundary data
is a trace of some functionbelonging
toWl,2(Q).
This condi- tion is ratherrestrictive,
because not every function inL 2
is thetrace of some function
belonging
to(see
Lions andMage-
nes
[17]
Theorems 7.5 and 9.4Chapter 1).
It is clear that the Dirichletproblem
withL2-boundary
datarequires
a new definition. The firstattempt
to define the Dirichletproblem
withL2 -boundary
data hasbeen made
by
Mikhailov who in a series of articles[11], [18], [19]
and
[20]
examined thisproblem
in a bounded domain under the assump- tion aij ECI(D) and bi,
c E(see
also Necas[22]
and[23]).
Similarresults were also obtained
by Kapanadze [15].
The author andThomp-
son
[5]
extended Mikhailov’s results to theequation
with coefficientssatisfying
somegeneral integrability
conditions. In the articles men- tioned above theboundary
8Qbelongs
to C2. For theLaplace equation
the Dirichlet
problem
withL2-boundary
data was solved in boundedLipschitz
domains(see Dahlberg [8],
Jerrison andKenig [12], [13]).
We mention here that Jerrison and
Kenig
also extended their results to bounded non-smooth domains for anequation
with C’-eoefficients(see [14]).
The
plan
of this paper is as follows. In sections 1-6 we examine the Dirichletproblem
in a half space. The main result is an energy estimate(section 3,
theinequality (19))
for theequation
~u+
lu =f,
where I is a
sufficiently large parameter.
Nextapplying
the resultof Bottaro and Marina
[3]
we establish the existence of aunique
solu-tion to the Dirichlet
problem
withL2-boundary
data for theequation (1)
with the conditione(x) ~
Const > 0. In sections7,
8 and 9 similarresults are established for the exterior Dirichlet
problem.
Methodsused in both cases are similar and
presented
in some details. Wepoint
out that theassumptions
on the coefficients ai~ for the Dirichletproblem
in the half space areconsiderably
weaker than thecorrespond- ing assumptions
for the exterior Dirichletproblem.
This follows fromthe fact that the
boundary
of the half space is flat.Among
the papers devoted tostudy
theboundary
valueproblems
in unbounded domainswe mention the work of Benci and Fortunato
[2],
in which the Dirichletproblem
with zeroboundary
data in aweighted
Sobolev space has been studied.In this paper we make
frequent
use of the Sobolevinequality
where
The most
significant
feature of thisinequality
lies in theindependence
of the constant S of the domain S~ which makes
possible
to use it inunbounded domains
(see
Federer[9]).
2. Traces.
Let
Rt
= >01.
We denote apoint
x ERi by x
== (x’,xn),
where x’= ... , IThroughout
sections 2-6 we make thefollowing assumptions
aboutthe
operator
L :(A)
.L isuniformly elliptic
in.Rn , i.e.,
there exists apositive constant y
such thatfor all
and $ e Rn ,
moreover(i, j = 1, ..., n).
(B) (i)
There existpositive
constants x and 0 a 1 such thatfor all and all
oo) .
All constants in this paper will be denoted
by C~ .
The statement«
C, depends
on the structure of theoperator
1)> means thatCi depends
on n, a,
b, k,
m, and the norms of the coefficientsDaij,
ai;,bi
andc in
appropriate
spaces. coIn the
sequel
we shall need thefollowing elementary
lemmas.is bounded
independently of 6
E(0, T/2].
PROOF.
Integrating by parts
Denoting
the lastintegral by
J andapplying Young’s inequality
we obtain
Taking s
=(1
-y)/4
the result follows.We also need the
following simple
observation.LEMMA 3. let be a solution
of (1 ).
Thenfor
everywhere a
positive
constant Cdepends
on the structureof
theoperator
Land r.PROOF. Let v =
U02
where 0 EUsing
v as a test function in( 2 )
we obtainIt follows from
ellipticity
of L and theinequalities
ofYoung
andSobolev that
where a
positive
constant Cdepends
on the structure of theoperator
JD. Here we have used the fact that c = c1+
C2 with Cl E J~ and c, ETo
complete
theproof put 0 = 0, ,
where0,
is anincreasing
sequenceof
non-negative
functions in with thegradient
bounded inde-pendently
of v andconverging
as v --~ oo to anon-negative
functionf on
lit equal
to 1 for andvanishing
forTHEOREM 1. be a solution
of (1)
in~.
Then thefollowing
conditions areequivalent :
PROOF. Let
0 C 38o C 1.
Define anon-negative
functionq e C2 ([0, oo))
such that = xn for and21(x,,)
= 1 forWe may assume that
q(rn) > 6
for all and 06 60.
Let
where 0 is a
non-negative
function in Since forevery 6
C x~v( ~, xn)
has acompact support
in it follows from Lemma 5 thatv is an admissible test function in
(2),
henceDenote the
integrals
on the left hand side of(4) by J1,
...,Je.
It followsfrom the
ellipticity
of .~ thatBy Young’s inequality
where a
positive
constant01 depends
on y andSimilarly
where a
positive
constantC, depends
on11 bi ll,..
Nowaccording
tothe
assumption (B iii)
where °1 E and 02 E
By
Hölder’sinequality
where
1/2*
=1 /2 - 11n.
Nowby
Sobolev’sinequality
were 8
depends only
on n. Since we may assumewe obtain
by Young’s inequality
where a
positive
constant03 depends
on n, and y.By
Green’s formula we haveand
by
theassumption (B ii)
where a
positive
constantC,~ depends
on and xi .Integrating
by parts
By
theassumption (Bi)
we obtainwhere a
positive
constantC, depends
on x, y and From the lastinequality
we deduce thefollowing
estimate forJ.,
where a
positive
constant06 depends
on x, and. Inserting
the estimates(5)-(10)
into(4)
we obtainIf the condition
(I)
holds thenby
Lemma 1 theintegrals
are bounded
independently
of 6. Nowput 0 = 0,, where 0,
is anincreasing
sequence ofnon-negative
functions intending
to1 as v - oo with the
gradient
boundedindependently
of v.Letting
v - oo in
(11)
it follows from Lemma 3 thatThe
implication
o I ~ II » follows from Lemmas 1 and 3 and theLebesgue
convergence theorem.To show that « II => I » observe that
Now
by
Lemma 2 the condition(II) implies
that theintegrals
are bounded
independently
of 6.Repeating
theargument
from thestep
«I =>Tl)) the result follows.REMARK 1. It follows from the
proof
Theorem 1 that the condi- tion(II) implies:
As an immediate consequence we obtain , COROLLARY 1..Let u E be a solution
o f (1)
inthat one
of
the conditions(I)
or(II)
holds. Then there exists afunction
q E and a sequence
6,
- 0 as v - oo such thatfor
everyTHEOREM be a solution
o f (1)
inR:. Suppose
that one
of
th e conditions( I )
or( II )
holds. Then there exists afunction
~ E such that
for every Y E L2(Rn-1).
PROOF. Since
~)2
dx’ isbounded,
say andann(x’, 6)
~n-l
continuous and bounded it follows from
Corollary 1,
thatIt suffices to show the existence of the
and
v(x)
= 0elsewhere, where 77 oo))
isthe
function introduced in theproof
of Theorem 1.Taking v
as a test function in(2)
we obtainSince for every 0
y
1"ri "n
and Y has a
compact support,
theLebesgue
dominated convergence theoremimplies
thecontinuity
ofNow
By
theLebesgue
dominated convergence theorem and theprevious
part
of theproof,
theright
hand side of the lastinequality
tends to0 as b - 0 and this
completes
theproof.
Our next
objective
is to establish theL2-convergence
ofu( ~, ð)
to 99 as 6 - 0. To do this we first show that the norm of
u( ., ð)
converges to the norm of 99. The result then followsby
the uniformconvexity
of the space
L2.j
THEOREM 3. Let U E
IVI,2 (R,+,)
be a solutionof (1 ). Suppose
thatone
of
the conditions(I)
or(II)
holds. Then there exists afunction
99 E L2(Rn_1)
such thatPROOF. If Il E
1’1’ ~’Z~lLn~~
thenby
theargument
used in theproof
of Theorem 1 we obtain
where q
is the function introduced in theproof
of Theorem 1. Defineit is clear that . Thus
We shall show that
.and
(15)
Indeed
Here we have used the fact that u is a solution of
(1)
and theidentity (2)
with the test functionwhere 27
is the function introduced in theproof ’of
Theorem 1.To prove
(15)
observe that.
Using
the conditions(I), (II)
and Lemma 1 one can show that Weonly
restrict ourselves to the termJ,.
Integrating by parts
the termJ3
can be written in thefollowing
formHence
by
theassumption (B ii)
where a
positive
constant Cdepends
on xl, n and(~== 11
...In -1). By
Lemma 1 the firstintegral
on theright
hand side tends to 0 as 6 -~ 0. Nowby
H61der’sinequality
It
follows
from the condition(I)
thatand
consequently by
the condition(II)
the secondintegral
on theright
hand side of(16)
also tends to 0 as 6 - 0. It follows from(~.3),
(14)
and(15)
thatTo
complete
theproof
observe that the normsare
equivalent
in3.
Energy
estimate.Consider the
elliptic equation
of the formin where A is a real
parameter.
The results of section 2
suggest
thefollowing
definitian of the Dirichletproblem.
A weak solution
is a solution of the Dirichlet
problem
with theboundary
conditionThe main result of this section is the
following
energy estimate.THEOREM 4. Let I be a solution
of
the Dirichletproblems
(17), (18).
Then there existpositive
constantsd, Ao
and Cindependent
of
’11, such thatfor all
A>
where apositive
constant Cdepends
on the structure of L.PROOF. Since is a solution of the Dirichlet
problem
for certain
30
> 0 andconsequently by
Theorem 1 the condition(II)
holds. All constantsappearing
inthe
proof
areindepenent
of 8 anddepend
on the structure of L and may vary from line to line.By
the same considerations as in theproof
of the
step
I => II » of Theorem 1 we show that~o, whereq
is the function introduced in theproof
of Theorem 1.Letting 6
- 0 we deduce from the lastinequality
It
follows
from(20)
and(3) (see
Lemma3)
thatNow
adding
theinequalities (3)
and(21)
weget
By
a similar manner(see
theproof
of thestep
« I => II » of Theorem1)
we obtain
Now
inserting (22)
into(23)
we obtainNow we make use of the
following
fact: for every0 ~ ~O
1 and 0d
1Thus
(19)
follows from(22)
and(23) provided A
issufficiently large
and d
sufficiently
small.. Dirichlet
problem
inRi.
We are now in a
position
to establish the existence of a solutionto the Dirichlet
problem.
THEOREM
~, ~ ~,o
Assume that r1(i =1,
..., n)
and that Thenfor (Rn_i )
thereexists a
unique
solutionof
the Dirichletproblem (17), (18)
inPROOF. Let
(qm)
be a sequence of functions inconverging
in
L2(Rn_i)
to the function 99. Put~rn =
1, 2, ....
. It follows from[3]
that there exists aunique
solutionin
>
of the Dirichletproblem
By
Theorem 4for where C is a
positive
constantindependent
of p and q. Con-sequently
is theCauchy
sequence in the normand the result follows.
THEOREM 6.
Suppose
that theessumptions of
Theorem 5 hold andmoreover 0 Then
for
thereexists a
unique
solution to the Dirichletproblem (1), (18)
inand moreover
where C is a
positive
constantdepending
on the structureof
theoperator
L.PROOF. Let
lo
be asufficiently large positive
constant.By
Theorem5 the Dirichlet
problem
has a
unique
solution inloc n satisfying
the energy estimate(25).
On the other hand in virtue of Theorem 1.1 in
[3]
the Dirichletproblem
has a
unique
solution in~W1~2(.Rn )
and moreoverwhere C is a
positive
constant. Thus the function u = ~o-f- v
is a solution of theproblem (1), (18)
and it is obvious thatwhere C is a
positive
constant. To obtain the estimate(25)
it sufficesto derive the estimate of the form
(23)
for sup~)2
dx’ and thenRn_1
use the final
part
of theargument
of theproof
of Theorem 4 and theinequality (27).
The next result establishes the relation between the solution of the Dirichlet
problem
in andT~1’2(.Rn).
Wepoint
out herethat
by
a solution of the Dirichletproblem
in we mean asolution of
(1)
with theboundary
condition in the sense of trace(see Introduction).
THEOREM 7.
Suppose
that theassumptions of
Theorem 6hold,
thatand that there exists a
f unction
such thatq;l(X’, 0 )
_~p(x’ )
onRn-l in
the senseo f
trace. Then the solutiono f
theDirichlet
problem (1), (18)
in solutionof
the sameproblem
in
PROOF. It follows from
[3]
that the Dirichletproblem
Zu =f
on
.Rn
and u = y onRn-l
has a solution in which is also asolution in The result follows from the
uniqueness
of thesolutions in of the
problem (1), (18).
5.
Weighted
estimate of thegradient.
It is well known that a solution of the Dirichlet
problem
for theLaplace equation
where q;
EL2(Rn-t)
satisfies thefollowing
relation-"n
The
question
arises whether theinequality,
RTn
remains true for the solution of the
problem (1), (18)
inc (see [26]
p.82-83).
Thefollowing
theorem containsa
partial
answer to thisquestion.
THEOREM 8.
Suppose
thatc(x) > Const
> 0 orc 7 c E+ + bi
= 0(i
=1,
...,n)
andf
= 0 onX -
Let u be a solutionof
the Dirichletproblem (1), (18)
in Thenwhere a
positive
constant Cdepends
on the structureof
theoperator
L.PROOF. Let be an
increasing
sequence of functions inC2([o, oo))
converging
to rn on[0, c>o)
withproperties: i7,(x,,)
= xn forq(rn) > 3
for xn >2 60
and~D2~v ~
are boundedindepen- dently
of v. q, may be chosen to be constant forlarge
Xnè i,Taking
as a test function in
(2)
we obtainWe denote the sum of the second and third
integral by J1
andintegrat-
ing by parts
we obtainwhere a
positive
constantC, depends
on m.,, supID*17,1
andsup
ID2,7,1. Similarly
the fourthintegral Jg
can be estimated in thefollowing
waywhere a
positive
constantO2 depends
on "1’ andsupID77,1.
+Finally using
theellipticity
and the fact that c is bounded belowby
.a
positive
constant we obtainetting
v - oo and 6 - 0 we derive from the lastinequality
where
positive
constantsC3
andÕ3 depend
on the structure of the opera- tor .~. Nowapplying
the energy estimate(25)
we obtain(28).
REMARK 2. The estimate
(28)
can be extended to the nonhomo-I
geneous
equation provided
1
6. Estimate of derivatives of’ the second order.
In this section we
replace
theassumptions (B i)
and(B ii) by
thefollowing
condition( k, i, j,
=1,...,’n),
whereb and x1 are positive
constants.
THEoRF,3i 9. u be a solution in
loc of
theproblem (1), (18).
T hen
where C is a
positive
constantdepending
on the structureo f
L.PROOF. It follows from Theorem 8.8 in
[10] (p.173) that u
EWe shall first show that for every r > 0
Indeed, taking v
=Dkw,
where w E withcompact support,
as a test function in
(2)
andintegrating by parts
we obtainNow let w = where 0 is a
non-negative
function in such that W = 0 for andxERn-l.
Then
Applying
theinequalities
ofH61der, Young
and Sobolev we deduce from the lastequation
thatwhere a
positive
constant Cdepends
on the structure of L. Nowput
where
(Py
is anincreasing
sequence ofnon-negative
functionsin with
supports
in onand with bounded
independently
of v.Letting
v - 00 the ine-quality (30)
follows. To establish(29)
letwhere 77
is a function introduced in theproof
of Theorem 1.By (30)
~,u is an admissible test function in
(31)
andWe may assume
Applying
theinequalities
ofYoung,
Holder and Sobolev we
easily
derive from the lastinequality
the follow-ing
estimatewhere C is a
positive
constantdepending
on the structure of the opera- torL,
and the result follows.7.
Exterior Dirichlet problem.
Preliminaries.Let Q c
Rn
be acomplement
of a bounded closed set withboundary
8Q of class C2. we consider the
equation (1).
Let x EQ,
we denoteby r(x)
the distance from x to 3D.We make the
following assumptions
There exists a
positive
constant y such thatfor all x E .~
and ~
moreover aij EL’(92)
r1(i, j
=1,
...,n).
We also assume that
xr(x)-a (i, j
=1,
... ,n)
in someneigh-
bourhood N of the
boundary 8Q,
where x > 0 and 0 a 1 are cons- tants and thatDkaii N) (k, i, j
=1,
...,n).
(~2) bi
Ef
E22(Q)
and c E+ Ln(,Q) (i
=1,
...,n).
It follows from the
regularity
of theboundary
8Q that there isa number
~o > 0
such that the domainwith
boundary
possesses thefollowing property:
to each
ro e 8Q
there is aunique point ro(so)
= xo - wherev(xo)
is the outward normal to 8Q at xo .According
to Lemma 1 in[10]
p.382,
the distancer(x) belongs
toC2(Q -
if6,
issufficiently
small. We extend
r(x)
aspositive
function of classC2(Q)
anddenote this extension
again by r(x).
We may assume that N c Q -Do..
Let xa denote an
arbitrary point
of8Qd.
For fixed 6 E(0,
letand
where
IA I
denote the(n - 1)-dimensional
Hausdorff measure ofA,
and
vo(xo)
is the outward normal to at xo. Mikhailov[19] proved
that there is a
positive
number Yo such thatWe will use the surface
integral
where u E and the values of
son
are understood in the sense of trace.Let
Ro
be apositive
number such thatQo
c.R~
forevery R >
Ro .
Set°
for all
..R > .Ro .
In the
sequel
we shall need thefollowing
estimate: ifand p
is a constant in[0,1),
thenfor 6 E
(0, ðo/4],
where .~1 is apositive
constantindependent
of 6 and u.This
inequality
can be establishedby
the sameargument
as in theproof
of Lemma
2;
in theproof
we use theinequality (32).
Moreover it is easy to see that if
M(6)
is bounded on(0, ~o]
thenfor every p E
[0,1)
for all 6 E
(0, bo/21,
whereM1
is apositive
constantindependent
of~ and u.
THEOREM 10. Let u be a solution
of (1 ) belonging
toWlo~(SZ).
Thenthe
following
conditions areequivalent
PROOF. We
only
sketch theproof
because it is identical to that of Theorem 1 in[5].
Fix an .R > and let 0 be anon-negative
func-tion in such that = 1 for
lxl c.R
andØ(x)
= 0 forIxl
> 2R.Suppose
that holds. Thus u E)
forevery
>.Ro
and(34)
holds. It is clear thatis an admissible test function in
(2)
andBy
Green’s formula(see [21],
p.139)
we haveand
By
theellipticity
condition and theinequalities
ofYoung
and Sobolevwe derive from
(35), (36)
and(37)
thatNow choose q _
0,,
where(Wr)
is anincreasing
sequence of non-negative
functions inconverging
to 1 for withIDØ,I
bounded
independently
of v.Letting 6 - o
and v - oo the resulteasily
follows.Now suppose that the condition
(Ill)
holds.By (35)
and(36)
we have
Using (33)
and theinequalities
ofYoung,
Holder andSobolev,
itis easy to see that
M(6)
is bounded on(0, 5o].
Our next
objective
is to prove that u has a trace on aS~ inL2(ôQ).
We first state the
following preliminary result,
which is easy to prove(see
Theorems 2 and 3 of Section2).
THEOREM 11..Let be a solution
of (1). Assume
thatone
of
the conditions(Ii)
or(III)
holds. Then there exists afunction
such that
for
every g E_L2(aQ).
To prove that - 99 in
L2(ôQ)
we show thatand the result follows from the uniform
convexity
of~SE
_ _
Fix .R >
.Ro,
E(0, ~o]
we define themapping lJ2R
-~by
where
xa(x)
denotes the nearestpoint
to x on and =xa~2(x)
for each x E 3D. Moreover >
5/2
and x8 isuniformly Lipschitz
continuous. Note that if u E then
u(r’)
E for each..R >
Ro .
To prove
L2-convergence
of we shall need thefollowing
tech-nical lemmas
LEMMA 4.
I f rif
and then5.
If gEL2(QR)
andr!fEL2(QR)
and suppose thatis bounded on
(0, ~o],
then a.,with inner
product (norm)
denotedwith inner
product (norm)
denotedTHEOREM 12. Let u E
loc
be a solutionof (1 )
such that oneof
the conditions
(11)
or holds. Then there is afunction
99 c such thatu(xa)
converges to 99 EPROOF. The
proof
is similar to that of Theorem 4 and is therepeti-
tion of the
argument given
in the case of bounded domain(Theorem
4in
[~]) .
Since
B1.BB1
and11.112
areequivalent
it sufficient to show that thereis 99
EL:
2 such that limu(xi)
= 99 in 2By
uniformconvexity
of.~2
it suffices to show thatBy
Green’s theorem we find thatLet 0 be defined as in the
proof
of Theorem10,
thenConsequently
we obtainfor 6 E
(0, 30],
since = x on,SZa,2R
and = 1 onDR.
We showthat
and
since = on
aS2.
The relation
(38)
follows from Lemmas 4 and 5. To establish(39)
we use
(2)
with a test function vgiven by
thefollowing
formulaTheorem
12 justifies
thefollowing
definition of the Dirichletproblem.
A weak solution of
(1)
is a solutionof the Dirichlet
problem
with theboundary
conditionB.
Energy
estimate for the exterior Dirichletproblem.
Assume that the distance function
r(x)
is extended into S~ in sucha way that
on >
Ro),
where1~1
is apositive
constant.Let q~
E and consider thefollowing
Dirichletproblem
inwhere A is a real
parameter.
IfA> 10 (10 sufficiently large)
thenby
Theorem 6 in
[5]
there exists aunique
solution. Theboundary
condi-tions
(44)
is understood in the sense ofL2-convergence and by (43)
we mean that
Wlt2(QR)
for every function P E such that P =1 in someneighbourhood
ofIx
_ Rand P = 0 on Q -LEMMA 6.
Suppose
that rsatisfies
the condition(E)
and let u be asolutions in
loc ) of
the Dirichletproblem (42), (43)
and(44).
Thenthere exist
10
> 0(sufficiently large)
and d > 0(sufficiently small)
suchthat
for
allR > .Ro,
wherepositive
constantÂo,
d and Cdepend
on the struc-ture
of
theoperator
and acreindependent o f
R.PROOF. The energy estimate
(45)
wasessentially proved
in[5].
We
repeat
theproof
to show that the constants0,
d andAo
areindepen-
dent of R. Let
on
elsewhere .
By
Green’s theoremNow
applying Holder
yYoung
and Sobolevinequalities
we obtain-where a
positive
constantCi
isindependent
of .Z~. On the other handwe have
Letting 8 -
0 in(47)
we obtainIt follows from
(48)
and(49)
that-where
O2
and03
arepositive
constantsindependent
of .R. Now observe thatby
theassumption (E)
we havewhere and
where and
consequently
Similarly
Choosing
dsufficiently
small andAo sufficiently large
the result follows from(48), (49), (50), (51)
and(52).
9. Existence of a solution of the exterior Dirichlet
problem.
It is clear that one can deduce from Lemma 6 the existence of a solution of the Dirichlet
problem (42), (43)
and(44)
in-Wl,2
As an immediate consequence of Lemma 6 we obtain
THEOREM 13.
Suppose
that the distancef unction
rsatisfies
the condi-tion
(E) . Let
99 E andf (x)2r(x)
dx oo. Thenfor
everyÀ;;;. Ào
there exists ac
unique
solutionof
the Dirichletproblem (42), (40)
and(41)
in and moreover
where a
positive
constant 0depends
on the structureof
theoperator
L.PROOF. For fixed .R >
.Ro
consider asolution vR
of theproblem (42), (43)
and(44)
and setvR(x)
= 0 forlxl
> .R. It follows from the energy estimate(45)
that there exists a sequence vRtending weakly
toa solution of the
problem (42), (40)
and(41).
From now on introduce the
following assumption
on the distancefunction r
(F)
The distance functionr(x)
on is extended to a function inC2(,~)
such thatr(x)
= 1 on S~ n~ .Ro~.
It is obvious that condition
(F) implies (E).
THEOREM 14. In addition to the
hypotheses
and(A2)
assumethat
bi
E(i
=1,
...,n),
c E+
and c > Const. > 0on Q. let and Then there exists a
unique
solu-tion u
of
the Dirichletproblem (1 ), (40)
and(41)
in and moreoverwhere a
positive
constant Cdepends
on the structureo f
Z.The
proof
is similar to that of Theorem 6 and therefore is omitted.We
only point
out that weagain
use here the results of Bottaro and Marina[3].
Under our
assumption
a solution in of theproblem (1), (40)-(41) belongs
toBy
the sameargument
as in theproof
of Theorem 9 one can establish the
following
estimate of the derivativeof the second order of a solution in of the
problem (1), (40)
and
(41).
THEOREM 15.
Suppose
that theassumptions of
Theorem 14 hold.Let ’U be a solution in
of
theproblem (1), (40)
and(41).
Thenwhere C is a
positive
constantdepending
on the structureof
theoperator
~.We mention here that the estimate of the derivative of the second order of a solution of the exterior Dirichlet
problem
in~2~2(S~)
wasobtained
by
Acanfora[1].
10. Case
c ~ 0.
To establish the existence and the
uniqueness
of a solution of theDirichlet
problem
we have assumed that c ~ Const > 0. If the coeffi- cient c isonly non-negative
one can also construct a solution butbelonging
to a different function space.Namely
introduce the space~1,2(Rn ) equipped
with the normand
similarly
for the exterior Dirichletproblem
the spaceJf’1,2(Q)
with the norm
where r s atisfies the condition
(.F’). By
Theorem 7 a solution of theDirichlet
problem (1), (18) belongs
andby
Theorem 14 asolution of the exterior Dirichlet
problem (1), (40)
and(41) belongs
to
y~l~ 2(,~)
.Now denote
by (D(Q))
thecompletion
ofwith
respect
to the normBy
Sobolev’sinequality (D(S2) c ~2*(S~)),
whencec
(-D(S2)
cTHEOREM 16.
Suppose
that theassumptions (A)
and(B )
hold. Letf
e.L2 (.I~ ) a’nd lfJ
e and moreover assume thatb
e(i
=1, ..., n),
and on Then there exists a sol’U- t’ion ’U to the Dirichletproblem (1), (18), belonging to the
spaceW1,2(R+n) +
+
Here the
boundary
condition(18)
is satisfied in thefollowing
sense:
for
every .R
> 0PROOF. Consider the Dirichlet
problem
where the
boundary
condition is understood inthe sense of L2-conver- gence(see
section3).
IfAo
issufficiently large
thenby
Theorem 5there exists a
unique
solution uobelonging
to the spaceT~1’2(.Rn ).
On the other hand
by
the result of Chicco[7]
the Dirichletproblem
has a solution in
D(X).
It is clear that v+
uo is a solution of(1), (18)
inT~’1’2(.Rn ) -~- D(.Rn ). The L2-convergence
in the sense(54)
followsfrom the fact that
In a similar manner one can establish
THEOREM 17.
Suppose
that theassumptions (A1)
and(A2)
hold..Let 99
E and moreover assume thatbi
E(i =1,
...,n),
c E
Ln/2(,Q)
andc(x) ~
0 on 92. Then there exists a solutionof
the Dirichletproblem (1), (40)
and(41)
in+ D(SZ).
REFERENCES
[1] F. ACANFORA, Problema esterno di Dirichlet per una classe di
equazioni
ellittiche del secondo ordine, Rend. Accad. Sei. Fis. Mat.
Napoli,
65(1979),
pp. 59-69.
[2] V. BENCI - D. FORTUNATO, Second order
elliptic
operators on unboundeddomains,
Boll. Un. Mat. Ital., (5), 15 (1978), pp. 193-209.[3]
G. BOTTARO - M. E. MARINA, Problema di Dirichlet perequazioni
ellittichedi
tipo
variazionale su insiemi non limitati, Boll. Un. Mat. Ital., (4), 8 (1973), pp. 46-56.[4] J. CHABROWSKI - B. THOMPSON, On traces
of
solutionsof a
semi-linearpartial differential equations of elliptic
type, to appear in Ann. Polon.Math.
[5] J. CHABROWSKI - B. THOMPSON, On the
boundary
valuesof
solutionsof
linear
elliptic equations,
Bull. Austral. Math. Soc., 27 (1983), pp. 1-30.[6] J. CHABROWSKI, Note on
nontangential boundary
limitsof
solutionsof elliptic equations in divergence form,
to appear inManuseripta
Math.[7] M. CHICCO, An
apriori inequality
concerningelliptic,
second orderpartial differential equations of
variational type, Matematiche(Catania),
26 (1971),pp. 172-182.
[8] B. E. J. DAHLBERG, On the Poisson
integral for Lipschitz
and C1-domains, Studia Math., 64 (1979), pp. 13-24.[9] H. FEDERER - W. H. FLEMING, Normal and
integral
currents, Ann. of Math., 72 (1960), pp. 458-520.[10] D. GILBARG - N. S. TRUDINGER,
Elliptic partial differential equations of
second order,
Springer-Verlag,
Berlin -Heidelberg -
New York, 1977.[11] A. K. GUSCIN - V. P. MIKHAILOV, On the
boundary
values in L-(p >
1)of
solutionsof elliptic equations,
Mat. Sb., 108 (150) (1979), pp. 1-19.[12]
D. S. JERISON - C. E. KENIG,Boundary
behaviourof
harmonicfunctions
in
non-tangentially
accessible domains,preprint,
Department of Mathe- matics, PrincetonUniversity.
[13] D. S. JERISON - C. E. KENIG,
Boundary
valueproblems
inLipschitz
do- mains,preprint.
[14] D. S. JERISON - C. E. KENIG, The Dirichlet
problem
in nonsmooth do- mains, Ann. of Math., 13 (1981), pp. 367-382.[15] J. V. KAPANADZE, On
boundary
valuesof a
second order linearelliptic equations,
Bulletin of theAcademy
of Science of theGeorgian
S.S.R., 93 (1979), pp. 285-288.[16]
O. A. LADYZHENSKAJA - N. N. URAL’TSEVA, Linear andquasilinear elliptic equations,
Academic Press, New York and London, 1968.[17] J. LIONS - E. MAGENES, Problèmes aux limites non
homogenes
etappli-
cations, vol. 1, Dunod, Paris, 1968.[18] V. P. MIKHAILOV,
Boundary
valuesof
solutionsof elliptic equations
inthe ball, Mat. Sb., 100 (142) (1972), pp. 1-13.
[19]
V. P. MIKHAILOV, Dirichlet’sproblem for a
second orderelliptic equation, Differential’nye Uravnenija,
12 (1976), pp. 1877-1891.[20] V. P. MIKHAILOV, On the
boundary
valuesof
the solutionsof elliptic
equa- tions,Appl.
Mat.Optim.,
6 (1980), pp. 193-199.[21] V. P. MIKHAILOV, Partial
differential
equations, Mir Publishers, Moscow, 1978.[22] J.
NE010DAS,
On theregularity of
second orderelliptic partial differential equations
with unbounded Dirichletintegral,
Arch. Rational Mech.Anal.,
9 (1962), pp. 134-144.[23]
J. NECAS, Les methodes directes en théorie deséquations elliptiques,
Aca- demia,Éditeurs, Prague,
1967.[24]
G. STAMPACCHIA, Leproblème
de Dirichlet pour leséquations elliptiques
du second ordre à
coefficients
discontinus, Ann. Inst. Fourier(Grenoble),
15 (1969), pp. 189-258.[25] G. STAMPACCHIA,
Équations elliptiques
du second ordre àcoefficients
dis-continus, Les Presses de l’Université de Montréal, Séminaire de Mathé-
tiques Supérieurs
(1965).[26]
E. STEIN,Singular integrals
anddifferentiability properties of functions,
Princeton
University
Press, Princeton, NewJersey,
1970.Manoscritto